considering multiple criteria for social categorization
TRANSCRIPT
Considering multiple criteria for social categorization can reduce intergroup bias Article
Accepted Version
final pre-publication version
Hall, N. R. and Crisp, R. J. (2005) Considering multiple criteriafor social categorization can reduce intergroup bias. Personality and Social Psychology Bulletin, 31 (10). pp. 1435-1444. ISSN 0146-1672 doi: https://doi.org/10.1177/0146167205276084 Available at https://centaur.reading.ac.uk/4486/
It is advisable to refer to the publisher’s version if you intend to cite from the work. See Guidance on citing .Published version at: http://dx.doi.org/10.1177/0146167205276084 To link to this article DOI: http://dx.doi.org/10.1177/0146167205276084
Publisher: SAGE Publications
All outputs in CentAUR are protected by Intellectual Property Rights law, including copyright law. Copyright and IPR is retained by the creators or other copyright holders. Terms and conditions for use of this material are defined in the End User Agreement .
www.reading.ac.uk/centaur
CentAUR
Central Archive at the University of Reading
Reading’s research outputs online
1
This is the final pre-publication version of this article including all responses to referees comments.
The definitive version of this article can be found here DOI: 10.1177/0146167205276084
MS PSPB-03-075 Version 4
Running Head: MULTIPLE CATEGORIZATION
Considering multiple criteria for social categorization can reduce intergroup bias
Natalie R. Hall & Richard J. Crisp
University of Birmingham
Word count: 7237
Author Note
Natalie R. Hall, Richard J. Crisp, School of Psychology, University of Birmingham.
This research was funded by a British Academy research grant (SG-34045) to R. J. Crisp and
an Economic and Social Research Council doctoral studentship (PTA-030-2002-00889) to N. R.
Hall.
Correspondence concerning this article should be addressed to N.R. Hall or R. J. Crisp at the
School of Psychology, University of Birmingham, Edgbaston, Birmingham, B15 2TT. Tel: +44
(0)121 414 2227 or +44 (0)121 414 3335. Fax: +44 (0)121 414 4897. E-mail: [email protected] or
Multiple categorization 2
Abstract
Two experiments tested the notion that considering multiple criteria for social categorization
can reduce intergroup bias. In both experiments, participants were required to consider alternative
ways in which people could be classified, other than an initially salient intergroup dichotomy. In
Experiment 1 we found that generating alternative social classifications that were unrelated to an
initial target dichotomy reduced intergroup bias compared to a control condition. In Experiment 2
we replicated this effect and found that unrelated, but not related, categorizations were necessary to
reduce bias. This paper adds support to the view that increasing categorical complexity is a useful
tool in bias reduction. We discuss these findings in the context of a developing model of multiple
categorization effects.
Keywords: INTERGROUP BIAS, MULTIPLE SOCIAL CATEGORIZATION, IDENTITY
COMPLEXITY
Considering multiple criteria for social categorization can reduce intergroup bias
Multiple categorization 3
Processing resources are a valuable commodity for the social perceiver. For the most part,
classifying others into broad groupings -- social categorization -- is a useful shortcut for navigating
the social world and thus saving such resources (Fiske & Taylor, 1991; Hamilton, 1979; Macrae &
Bodenhausen, 2001). The usefulness of this mechanism can be demonstrated by the observation that
it is used even in artificial and arbitrary contexts (e.g., Rabbie & Horowitz, 1969; Tajfel, 1982;
Tajfel, Flament, Billig & Bundy, 1971; for a review see Bourhis, Sachdev, & Gagnon 1994). The
cognitive efficacy of social categorization comes, however, with some drawbacks: it is the mental
pre-requisite for intergroup bias. People in the perceiver’s own groups tend to be reliably and
consistently evaluated more positively than those in other groups (e.g., Brewer, 1979, Mullen,
Brown & Smith, 1992). In this article, we explore the potential for reducing such bias by
contravening the inherent ‘functionality’ of categorization.
The usefulness of social categorization
Categorization causes an accentuation of differences between, and similarities within, social
groups (Doise, 1978; McGarty & Penny, 1988; Tajfel & Wilkes, 1963). According to many theorists,
it is this process of differentiation between groups that, in part, leads to ingroup favouritism (for
examples see the Category Differentiation Model, Doise, 1978; Social Identity Theory, Tajfel &
Turner, 1979; Self-Categorization Theory, Turner, Hogg, Oakes, Reicher & Wetherall, 1987). There
are differences between these accounts. For some categorization is a simplifying mechanism, for
others a way of ascribing meaning to the social context, but they all rest on the notion that to
observe intergroup bias there must first exist a distinct cognitive dichotomization into ‘us’ and
‘them’ (see Maass & Schaller, 1989; Tajfel, 1969), and they are all linked by a common principle:
that categorization is on some level psychologically ‘useful’.
Multiple categorization 4
Social categorization can be considered useful because it is adaptive, providing a quick and
sufficient understanding of what is occurring in the social world (van Knippenberg & Dijksterhuis,
2000). Group-based judgments represent a slice of knowledge that can be applied to a target person
who fits the category (e.g., Sherman, Macrae, & Bodenhausen, 2000). Categorical thinking is
cognitively economical -- providing the maximum output of information with the least cognitive
effort (Macrae & Bodenhausen, 2001). It reduces the need to process incoming data by providing the
perceiver with group-based expectancies; and directs attention and processing resources to
expectancy relevant information (Hastie, 1981; Stangor & Macmillian, 1992). Using social categories
to guide processing therefore reduces the cognitive resources required for person perception by
providing a set of pre-conceived expectations about the target (Bodenhausen, 1992; Macrae, Stangor
& Milne, 1994; Oakes & Turner, 1990).
Although categorical thinking can be advantageous from this functional perspective,
stereotypes associated with category activation are often inaccurate when applied to individual
targets. Furthermore, such activated social categories can cause selective perception, interpretation,
and memory (Heilman, 1995). This can be problematic when the (sometimes negative) stereotypic
information generated is used to guide impression formation. The question we addressed in this
research was whether removing the cognitive efficacy of categorization would prevent its use in
guiding evaluative judgments in person perception. We suggest that one way to achieve this is by
making multiple alternative bases for social categorization salient for the perceiver.
Multiple social categorization
A considerable amount of work has focused on the notion that changing the categorical
representation of groups can correspondingly change intergroup evaluations. Levine and Campbell
(1972) drew attention to the reduced conflict associated with cultures having a crossed societal
Multiple categorization 5
structure. They noted that in contexts characterized by social structures that comprise loyalties to
multiple groups, some of which coincide, and some of which are in conflict, lower levels of
intergroup conflict tend to be observed. Later, this idea emerged in social psychological work on
intergroup relations. Deschamps and Doise (1978) suggested that by simultaneously combining two
cross-cutting dimensions of social categorization, opposing accentuation and assimilation principles
would effectively cancel each other out and reduce intergroup bias. This approach of combining two
cross-cutting categorizations has met with mixed success in terms of reducing intergroup bias
(Brown & Turner, 1979; Crisp, Hewstone, & Rubin, 2001; Deschamps & Doise, 1978;
Vanbeselaere, 1987; 1996). The majority of studies produce a pattern of evaluation where any group
associated with the ingroup is evaluated positively to some extent, but where negative evaluations
are still directed towards those people who are outgroup according to both categorical criteria
(Diehl, 1990; Ensari & Miller, 1998; Hewstone, Islam & Judd, 1993; Kenworthy, Canales, Weaver,
& Miller, 2003; Marcus-Newhall, Miller, Holtz & Brewer, 1993; Urada & Miller, 2000;
Vanbeselaere, 1987; see Crisp & Hewstone, 1999; Urban & Miller, 1999 for reviews).
In contrast to focusing participants on two cross-cutting dimensions of categorization
(Deschamps & Doise, 1978; Vanbeselaere, 1987; 1996), considering more than two categorization
criteria may uniquely lead to a ‘decategorization’ effect (Crisp et al., 2001). The reasoning is as
follows: we know that participants are cognitively able to use two dimensions of social
categorization under normal processing conditions (e.g., Vanbeselaere, 1987). As the number of
categorization dimensions to consider increases, the intergroup context would become much more
complex and the perceiver may no longer be able to easily determine on what basis they are similar
to, or different from, the target person. There may be too many dimensions of membership to allow
meaningful category conjunctions to form. Perceivers may be able to cognitively combine categories
Multiple categorization 6
like ‘single mother’ or ‘working parent’ (see Hutter & Crisp, in press), but they may find, for
example, ‘single, black, educated, working mother’ more difficult, at least with respect to gaining
any added value by processing information and making evaluative judgments heuristically. There
may come a point in considering successive categorical dimensions when using a heuristic method of
classification actually becomes inefficient relative to other methods of impression formation. The
idea of multiple modes of processing is key to contemporary theories of person perception, and we
suggest that the ideas outlined above regarding the effects of multiple categorization can be
interpreted within the context of some well-established frameworks relating to impression
formation.
Categorization and decategorization
Brewer’s (1988) Dual Process Theory and Fiske and Neuberg’s (1990) Continuum Model
both consider person perception and impression formation to comprise of two distinct processes:
categorization and individuation. The basic premise of Brewer’s Dual Process Theory (1988) is that
either a top-down category based approach or a bottom-up person based approach is used and this is
analogous to a heuristic (category) versus systematic (individuated) approach to person perception.
Fiske and Neuberg (1990) propose a continuum for impression formation where one extremity is
category-based processing and the other is attribute (individuated) based processing. On this
continuum people can be perceived as a representative of a group (category-based responding), or as
an individual separate from any categorical identity (attribute-based responding). Fiske and Neuberg
propose a system where an initial categorization will occur, followed by a confirmatory process to
determine how good a fit there is between the category and target. If this is unsuccessful, then there
will be a shift towards the individuated mode of perception by invoking an attribute-by-attribute
approach to form an impression of the target person. As Fiske, Lin and Neuberg (1999) state:
Multiple categorization 7
“...perceivers typically use category-oriented processes before they use attribute-orientated processes,
and [that] if the category-oriented processes work well enough perceivers do not engage additional
more attribute oriented processes” (pp. 236). Brewers’ Dual Process Theory is based on similar
principles and predicts that the mode of perception will change from the use of category based
heuristics to a more systematic individuated mode of perception under conditions that either favor,
or do not favor, one over the other. As Brewer and Harasty Feinstein (1999) note, the changes in
processing outlined by the Dual Process Model map directly onto Fiske and Neuberg’s impression
formation continuum.
This switch in processing from using categorization to individuation can be termed
decategorization. Decategorization results when initial categorizations of people as group members
are overridden by a more comprehensive and differentiated perception of the individual (see Brewer
& Miller, 1984; Ensari & Milller, 2001; Wilder, 1986). Once decategorization has occurred the target
person should be primarily defined as an individual rather than a group member, which should
remove category-based bias. Previous research has found that conditions that promote
decategorization will encourage participants to favour the ingroup less (presumably because the
judgment is being made on an appreciation of personal merit, not on the basis of pre-conceived
categorical expectancies; see Brewer & Miller, 1984; Krueger & Rothbart, 1988; Miller &
Harrington, 1992). This change in the mode of processing should also allow the perceiver to develop
a more personalized and less homogenous perception of ingroup and outgroup members (Ensari &
Miller, 2001; Fiske & Neuberg, 1989).
We argue that considering multiple criteria for social categorization could produce a
decategorization effect by virtue of the implied decrease in usefulness of any one dimension of social
comparison. Decategorization is, however, an uncommon finding when only two dimensions of
Multiple categorization 8
categorization are combined. Ensari and Miller (2001) directly compared personalized contact
(where participants are asked to consider unique attributes) with the traditional crossed
categorization approach (two dimensions of categorization). They found that intergroup bias was no
longer moderated by shared ingroup memberships in the personalised contact condition, but this
was not the case in the crossed categorization condition. This could be because participants are quite
able (and appear to) consider two social categorizations simultaneously in memory and use them to
guide evaluative judgments (Vanbeselaere, 1987; 1996).
However, as the number of category dimensions that need to be processed simultaneously
increases then it may be that categorization, as a functional heuristic, becomes relatively inefficient,
leading to a shift in orientation and a reduction in bias as outlined by Brewer’s (1988) and Fiske and
Neuberg’s (1990) models. If perceivers experience conditions in which categorization becomes too
complex to be functional then they may resort to a more individuated strategy for person
perception. Correspondingly they should be less biased in their evaluations of ingroup and outgroup
members, as these evaluations are no longer tied to group membership. By considering multiple
criteria for social categorization, category-oriented processes may no longer provide the most
functionally efficient or meaningful way of making judgments, and consequently perceivers may
resort to a less biased approach to person perception.
In the context of the dual processing accounts of impression formation outlined above, we
can make some clear predictions with respect to the effects of multiple categorization on intergroup
evaluations. We thus tested the possibility that merely thinking about multiple alternative criteria
for social categorization may decategorize person perception and reduce category-based ingroup
favoritism.
Pilot study
Multiple categorization 9
We initially carried out a pilot study to develop our multiple categorization task. We focused
on undergraduate students’ own university group membership versus a local rival university. Forty-
five undergraduate psychology students took part in individual sessions. All participants were
informed that they would be asked to think about people who shared their university membership
and people who did not. They were advised that the study was concerned with perceptions of
people at different universities and how their respective qualities influence creativity and originality
in working environments, and that all information would be compiled to assess the skills and
abilities from graduates at different universities. Participants first completed a task that was
ostensibly designed to test creativity, originality, and artistic merit. They were provided with a
square piece of paper and asked to make a model that represented something in the real world by
folding or tearing. They were allowed five minutes to complete this task. Participants in the
experimental condition were then given the following instructions that constituted the critical task
variable manipulation: “On an earlier page we asked you to think about your university affiliation.
This is one possible way that people can define themselves, but there are many more. People can be
similar or different from others not just according to one criterion, but often there are many and
varied ways that people define themselves and others. We would like you to think of other groups
that university students can use to define themselves (in addition to what university they are at)”.
They were asked to provide a few written lines of elaboration on each alternative they came
up with to reinforce the manipulation. They then completed a measure of group evaluation.
Participants in the control condition did not complete a categorization task. Instead, they
completed the group evaluations directly after making their paper model and received none of the
instructions outlined above.
Multiple categorization 10
Evaluations were measured by participants rating two paper models supposedly made by
previous ingroup and outgroup participants1
. Participants were asked to rate an ingrouper’s and
outgrouper’s paper model according to creativity, originality and artistic merit on scales ranging
from 0 (not at all) to 9 (very much so). These items were averaged to form indices of group
evaluation: ingroup =.721, outgroup =.722. The group to which the model was affiliated and the
order in which the models were presented was counterbalanced across conditions.
A 2 (task: control vs. multiple alternatives) x 2 (target: ingroup vs. outgroup) mixed
ANOVA with repeated measures on second factor revealed a significant interaction between task
and target F (1, 43) = 4.77, p = .035. This interaction reflected significant variation in intergroup
bias (ingroup minus outgroup evaluations) as a function of task instructions, although not in the
expected direction. Compared to the control condition (M= -.300), intergroup bias was increased
following the generation of multiple alternatives (M = .707)2
.
To investigate why we did not obtain the predicted reduction in bias we examined the
specific alternative classifications generated by our participants in the experimental condition. We
found that the majority of participants who completed the multiple categorization task (78%) had
actually produced classifications that were related in some way to the initial target dichotomy, rather
than unrelated alternative ways of classifying the student groups. We define related category
dimensions as those that are conceptually linked with the target categories. For example, what
academic subject [major] is being studied is related to the context within which the target intergroup
dichotomy (in this case ‘students’) is embedded. In contrast, unrelated category dimensions are those
that are not conceptually linked with the target categories. For example, in our studies ‘nationality’
is unrelated to being a student. In this pilot study most participants had generated category
dimensions like ‘subject [major] studied’ or ‘hall or residence [fraternity/sorority]’ as their
Multiple categorization 11
alternative criteria. These alternatives are, however, directly related to the dimension ‘student
affiliation’. Few participants generated alternatives that were (as we had intended) unrelated to the
target intergroup context (e.g., ‘gender’ or ‘place of birth’). Interestingly, closer examination of the
participants who had generated unrelated criteria did show a trend in the predicted direction
(reduced bias).
We hypothesized that generation of related alternatives may not undermine the usefulness of
the target dichotomy. Our reasoning was as follows. We know from previous work that whilst
making a superordinate categorization salient can sometimes reduce intergroup bias (Gaertner &
Dovido, 2000), sometimes it can be ineffective, or even increase bias (see Hornsey & Hogg, 2000). In
their Ingroup Projection Model Mummendy and Wenzel (1999) draw upon Self-Categorization
Theory (Turner, 1981; Turner, Hogg, Oakes, Reicher & Weatherall, 1987) to argue that making a
superordinate categorization salient can accentuate intergroup bias by introducing a relevant basis
for social comparison. For example, fans of opposing sports teams may not be concerned with
making ingroup favoring social comparisons against some other team until a relevant superordinate
dimension is salient (i.e., they become rivals within the same league or competition).
We argue that the generation of related alternative categorizations may have similar effects.
Thinking of related categorizations, by definition, should prime a relevant superordinate category
dimension (‘students’) along which to compare the initial ingroup and outgroup student subgroups
(defined by university affiliation). For example, categories related to university affiliation such as
fraternity or sorority, or major studied, all make the superordinate category ‘students’ more salient
and relevant. Put another way, thinking about bases for categorization other than the target
dichotomy, but nonetheless related to it, creates a context of comparability in which ingroup
favoring evaluations can serve to positively differentiate the ingroup from the outgroup (Tajfel &
Multiple categorization 12
Turner, 1979; see also Hogg & Abrams, 1990; Jetten, Spears & Manstead, 1996; 2001). Generation of
related alternatives reinforces any potential comparability between the ingroup and outgroup, thus,
ingroup favoritism should at best, remain pervasive, and at worse increase.
This pilot study therefore established an important pre-condition for the generation of
multiple alternative classifications to reduce intergroup bias. To avoid priming a superordinate
context of comparison as outlined above, it would be important to specify the generation of
unrelated alternatives. In our first experiment, we therefore modified our instructions to achieve this
goal.
Experiment 1
In this experiment the multiple categorization task instructions used in the pilot study were
adapted to encourage participants to generate unrelated, not related, alternatives. We predicted that
relative to a control condition, generation of unrelated alternatives would result in a decrease in
intergroup bias.
Method
Participants and Design
Thirty-eight undergraduate psychology students (31 females, 7 males, with an overall mean
age of 19) were randomly allocated to a 2 (task: control vs. multiple unrelated alternatives) x 2
(target: ingroup vs. outgroup) mixed design with repeated measures on the second factor.
Participants received course credit for their involvement.
Procedure
The procedure was identical to the pilot study except that after having created the paper
models, participants were given the multiple categorization instructions outlined above with the
addition of an explicit instruction not to generate related alternatives (the student subgroups ‘subject
Multiple categorization 13
studied ’ and ‘hall of residence’ were given as examples of alternatives to avoid). Participants were
asked instead to generate “alternative ways people can vary independent from being a student”, with
the example given of ‘Nationality’.
Participants in the control condition did not carry out the multiple categorization task and
simply completed the evaluative measures (the typical baseline condition used in bias-reduction
research) using the same paper models as employed in the pilot study.
Dependant Measures
Evaluations were measured by participants rating two paper models for creativity on a scale
ranging from 0 (not at all) to 9 (very much so)3
. One of the models ostensibly belonged to an
ingroup member and one ostensibly belonging to an outgroup member. The group to which the
model was affiliated and the order in which the models were presented was counterbalanced across
conditions.
Results & discussion
One participant did not fully complete the dependant measures so was excluded from all
further analysis. Four participants were removed on the grounds that they were suspicious about the
aims of the experiment. Despite the refined multiple categorization instructions one participant
exclusively produced related classifications so they were also excluded leaving an N of 32, with the
remaining participants evenly distributed across conditions. Means and standard deviations can be
found in Table 1.
Manipulation check
The categories generated by participants in the unrelated alternatives condition were judged
for relatedness by two independent raters blind to the experimental hypotheses (participants in the
control condition did not produce any categories). The judges were asked to rate the five category
Multiple categorization 14
dimensions produced by each participant on a seven point scale as to whether they were unrelated
(anchored at 1) or related (anchored at 7) to the initial target category dimension (student affiliation).
Judges were provided with an example to clarify this distinction. An example of an unrelated
category was ‘hair color’, an example of a related category was ‘subject studied’. The two blind
coders’ ratings had acceptable inter-rater reliability, r = .688, p =.002, (MJudge 1 = 1.41, SDJudge 1 = .618
vs. MJudge 2 = 2.12, SDJudge 2 = 1.50) so were collapsed into one scale (by taking the mean of the two
judges’ ratings). A one sample t-test against the scale midpoint (4) revealed that on aggregate
participants in the experimental condition had produced category dimensions that were unrelated to
the target dimension of comparison (‘students’), t (16) = -9.35, p < .0005 (M = 1.76, SD = .986).
Evaluations
A 2 (task) x 2 (target) mixed ANOVA with repeated measures on the second factor revealed no
main effect of target, F (1, 30) = .868, p = .359, nor of task, F (1, 30) = 2.41, p = .131. There was,
however, the predicted interaction between task and target, F (1, 30) = 7.57, p =.010. In line with
expectations, intergroup bias (ingroup minus outgroup evaluations) was lower following the
generation of multiple unrelated alternative categorizations (M = -.083) compared to the control
condition (M = 1.67)4
.
In summary, in line with predictions, intergroup bias was reduced following the generation of
multiple unrelated categorizations compared to the control condition. The effect appeared to be
driven predominantly by lowered ingroup evaluations, although some (non-significant) variation in
outgroup evaluations was also observed. We know that intergroup bias is driven primarily by
accentuated ingroup evaluations rather than outgroup derogation (Brewer, 1979). As such, reduced
reliance on social categorization should, as we observed, result in a decrease in this ingroup
favoritism (which is driving the bias in the first place), rather than a decrease in outgroup
Multiple categorization 15
derogation. In Experiment 2 we sought to replicate and expand on these findings by directly
comparing the effects of generating related versus unrelated alternative bases for categorization.
Experiment 2
Our pilot work indicated that thinking about related alternative classifications had a
detrimental effect on intergroup relations, yet, we did not experimentally test this in Experiment 1.
In Experiment 2 we therefore directly compared the two ways of framing the task. We had decided
to use only a single-item evaluative measure in Experiment 1 so we expanded this to a three-item
measure in this experiment. We expected reduced intergroup bias only for participants who were
instructed to generate unrelated alternatives, compared to a control condition and a condition in
which participants generated related alternatives.
Method
Participants and Design
Fifty-eight undergraduate psychology students (56 females; 2 males with a mean age of 19)
were randomly allocated to a 3 (task: control vs. multiple unrelated alternatives vs. multiple related
alternatives) x 2 (target: ingroup vs. outgroup) mixed design with repeated measures on the second
factor. Participants took part in exchange for course credit.
Procedure
The procedure was the same as in Experiment 1 except we added a related alternatives
condition. This version of the multiple categorization task comprised the same core instructions
used in the pilot study but instead of the addition of an explicit instruction not to generate related
alternatives, participants were specifically instructed to generate related alternatives (the student
Multiple categorization 16
subgroups ‘subject area’ and ‘hall of residence’ were given as examples of alternatives to use). The
‘unrelated’ task was identical to that used in Experiment 1.
Dependent measures
Participants were asked to rate the ingroup and outgroup paper models according to
creativity, originality and artistic merit on scales ranging from 0 (not at all) to 9 (very much so).
These items were averaged to form indices of ingroup ( = .762) and outgroup ( = .810) evaluation
respectively.
Results & discussion
Three participants were removed because they reported suspicion about the aims of the
experiment, leaving an N of 55. Means and standard deviations for all measures can be found in
Table 2.
Manipulation check
The categories generated by participants in the related and unrelated alternatives conditions
were judged by the same two independent raters, blind to condition, as described in Experiment 1.
We also used the same ‘relatedness’ scale for judging participants’ responses as used in Experiment 1.
The ratings produced by the two blind coders had acceptable inter-rater reliability, r = .905, p <
.001 (unrelated: MJudge 1 = 1.20, SDJudge 1 = .696 vs. MJudge 2 = 1.85, SDJudge 2 = 1.18, related: MJudge 1 =
4.44, SDJudge 1 = 1.58 vs. MJudge 2 = 5.56, SDJudge 2 = 1.89) so were collapsed (by taking the mean) into
one index. On this relatedness index, a t-test revealed a significant difference between related and
unrelated conditions in the expected direction, t (36) = -8.30, p < .0005, (Munrelated = 1.52 vs. Mrelated =
5.00). Participants in the related condition generated categories that were more related to the student
superordinate category context than those in the unrelated condition. One sample t-tests were
carried out against the midpoint of the scale. Those in the related condition produced categories that
Multiple categorization 17
were judged to be significantly related to the superordinate context, t (17) = 2.58 p = .020. In
contrast, those in the unrelated condition produced categories that were judged to be significantly
unrelated to the superordinate context, t (19) = -13.01, p < .0005.
Evaluations
A 3 (task) x 2 (target) mixed ANOVA with repeated measures on the second factor revealed
no main effect of target, F (1, 52) = .004, p = .950, nor of task, F (1, 52) = .261, p = .771. There was,
however, the expected significant interaction between task and target, F (2, 52) = 3.84, p = .028.5
To
explore the observed interaction, we again created an index of intergroup bias (ingroup minus
outgroup evaluations). Consistent with the findings from Experiment 1, planned comparisons
revealed less intergroup bias in the unrelated alternatives condition than in the control condition, t
(52) = 2.23 p = .030 (Mcontrol = .249, vs. Munrelated = -.650). There was also less intergroup bias in the
unrelated alternatives condition compared to the related alternatives condition, t (52) = 2.49, p =
.016 (Munrelated = -.650, vs. Mrelated = .338). Finally, there was no difference in intergroup bias between
the control and the related alternatives condition, t (52) = -.219, p = .828, (Mcontrol = .249, vs. Mrelated
= .338).
To summarize, in Experiment 2 we supported the idea that merely thinking about multiple
alternative categorizations can reduce intergroup bias -- as long as they are unrelated to the initial
intergroup context.
General Discussion
The aim of these experiments was to see if considering multiple alternative categorizations to
the target ingroup versus outgroup categorization could reduce intergroup bias. In Experiment 1 we
found that the generation of multiple unrelated alternatives led to the predicted bias-reduction
effect. Experiment 2 built on these findings by replicating the decrease in intergroup bias following
Multiple categorization 18
generation of multiple unrelated, but not related, categorization criteria. Overall, these studies have
developed and refined a new approach to bias-reduction that is consistent with extant theoretical
and empirical work on dual processing accounts of impression formation.
Multiple alternatives and decategorization
In Experiments 1 and 2 we found that the generation of multiple criteria for social
categorization that were unrelated to the superordinate category comparison context ‘students’
reduced intergroup bias. The findings are consistent with the suggestion that the generation of
multiple alternatives can encourage a shift in judgment orientation: if categorization is no longer
functional in such contexts then judgments become less category-based (Brewer, 1988; Fiske &
Neuberg, 1990), reducing ingroup favoritism.
There is an important caveat to this approach. Although the generation of multiple
categories can reduce ingroup favoritism, it will only do so if the categories generated are unrelated
to the superordinate category context. We believe that this is a crucial prerequisite for ensuring the
success of multiple categorization approaches that aim to achieve decategorization. The generation
of related categories may emphasize ingroup and outgroup comparability by highlighting the
superordinate category context. Correspondingly, this may have no bias-reducing effect, or even
result in a reactive increase in intergroup bias.
There are a number of potential explanations for what might cause the observed reduction in
bias. We have argued that the generation of multiple categories may undermine the utility of the
initial dichotomy and hence prevent the application of stereotypes relating to initial ingroups and
outgroups. It is also possible, however, that producing numerous categorizations may reduce bias
along the specified ingroup-outgroup categorization by reducing the degree to which the categorical
Multiple categorization 19
distinction is salient for the perceiver. The decategorization effect may also be caused by a dilution
of the original groups and their associated characteristics with many other categorizations and their
stereotypes. The aim of the current research was to establish whether generation of multiple
alternative categorizations can lead to a reduction in bias. Having now established this, more
research is needed to delineate precisely the mechanisms that underlie this effect.
We should acknowledge that some aspects of our findings should be considered tentative. We
have demonstrated the effect on only one dependent measure and with one paradigm thus, it is
necessary for future work to test generalizability of the findings. In addition, while the general
pattern of the findings across studies is consistent with predictions, there are also a number of
specific inconsistencies across the three studies (relating to whether the locus of effect resides in
moderation of ingroup evaluations, outgroup evaluations, or both). This further strengthens the call
for replication with other procedures and measures.
Category complexity and ingroup projection
The multiple categorization approach used in this research may relate to the Ingroup
Projection Model, Mummendy and Wenzel (1999). This model builds upon Self-Categorization
Theory (Turner, 1981; Turner et al., 1987) to argue that superordinate categorization creates a
context of comparability between ingroup and outgroup subgroups. Mummendey and Wenzel argue
that in intergroup situations where the superordinate group is salient, the ingroup prototype will be
projected on to the superordinate group. This ingroup projection may cause the outgroup to be
perceived more negatively. If both the ingroup and outgroup are conceptualized as part of a
superordinate category (a relevant comparison context), and the ingroup prototype has been
projected onto the superordinate category, then the outgroup will be seen as deviating from the
norm (Wenzel, Mummendey, Weber, & Waldzus, 2003). The ingroup projected prototype will
Multiple categorization 20
invariably be positive and hence the outgroup will be judged less positively by comparison (Wenzel,
2001). We argue that the generation of category dimensions related to the target dichotomy primes
the superordinate categorization context, making the ingroup and outgroup similarly comparable.
In contrast, generating unrelated alternatives reduces bias because it creates a context in which the
utility of categorization per se is reduced – the context becomes too complex to usefully employ
social categories.
Interestingly, the notion that category complexity can reduce intergroup bias is also reflected
in the Ingroup Projection Model. Waldzus, Mummendey, Wenzel, and Weber (2003) found that
projection of the ingroup prototype does not occur when the superordinate category is perceived
itself as having a complex prototype. Although we argue that generating unrelated alternatives
should lead to categorization being abandoned as a means of structuring social perception (so should
not invoke the superordinate category), there are clear parallels between our model and the Ingroup
Projection Model, in that both argue that complexity can reduce intergroup bias. Indeed, the notion
that complexity is a good thing for intergroup relations is emerging as a key theme in work on
intergroup relations. Roccas and Brewer’s (2002) model of Social Identity Complexity argues that
the more complex a perceiver’s social identity, and the less overlap there is between their multiple
ingroup identities, the less ingroup favoritism should be observed. Although not specific to
ingroups, our research supports this general concept.
The findings we report reiterate the message that multiple categorization can reduce
intergroup bias. Future research could examine some interesting possibilities regarding the
integration of multiple categorization and ingroup projection via notions of complexity, and the
corresponding potential for developing effective interventions to reduce intergroup bias.
Conclusions
Multiple categorization 21
The aim of this research was to explore the potential for multiple social categorization to
reduce intergroup bias. Could merely thinking about multiple alternative bases for social
categorization be an effective strategy for promoting more egalitarianism in person perception? On
the basis of these findings, we argue that the answer is yes -- but with one important caveat. The way
in which people are asked to consider multiple alternative criteria is critical to the observed effects.
We found that categorization no longer defined evaluative judgment following multiple
categorization providing that the categories considered were not related to the superordinate
category context. These findings provide some important links between diverse literatures in
impression formation and person perception, but perhaps more importantly, may also go some way
to developing an integrated model of the implications of multiple categorization for intergroup
relations.
References
Bodenhausen, G. V. (1992). Information processing functions of generic knowledge
structures and their role in context effects in social judgement. In N. Schwartz & S. Sudman (Eds.),
Context effects in social and psychological research (pp.267-278). New York: Springer Verlag.
Bourhis, R. Y., Sachdev, I., & Gagnon, A. (1994). Intergroup research with the Tajfel
matrices: Methodological notes. In M. P. Zanna & J. P. Olson (Eds.), The psychology of prejudice: The
Ontario symposium, (pp. 209-232). New York: Random House.
Brewer, M. B. (1979). Ingroup bias in the minimal intergroup situation: A cognitive-
motivational analysis. Psychological Bulletin, 86, 307-324.
Multiple categorization 22
Brewer, M.B. (1988). A dual process model of impression formation. In T. K. Srull & R. S.
Wyer Jr (Eds.) Advances in Social cognition (pp. 1-36). Erlbaum: Hillsdale, NJ.
Brewer, M. B., & Harasty Feinstein, A. S. (1999). Dual processes in the cognitive
representation of persons and social categories. In S. Chaiken & Y. Trope (Eds.) Dual process theories
in social psychology (pp. 255-270). New York: Guilford.
Brewer, M. B., & Miller, N. (1984). Beyond the contact hypothesis: Theoretical perspectives
on desegregation. In N. Miller & M. B. Brewer (Eds.), Groups in contact: The psychology of
desegregation (pp. 281-302). San Diego, CA: Academic Press.
Brown, R. J., & Turner, J. C. (1979). The criss-cross categorization effect in intergroup
discrimination. British Journal of Social & Clinical Psychology, 18, 371-383.
Crisp, R.J., & Hewstone, M. (1999). Differential evaluation of crossed category groups:
Patterns, processes, and reducing intergroup bias. Group Processes & Intergroup Relations, 2, 307-333.
Crisp, R. J., Hewstone, M., & Rubin, M. (2001). Does multiple categorization reduce
intergroup bias? Personality & Social Psychology Bulletin, 27, 76-89.
Deschamps, J. C., & Doise, W. (1978). Crossed category memberships in intergroup
relations. In H. Tajfel (Ed.), Differentiation between social groups (pp. 141-158). Cambridge:
Cambridge University Press.
Diehl, M. (1990). The minimal group paradigm: Theoretical explanations and empirical
findings. In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology, (Vol.1, pp. 263-
292). Chichester: Wiley.
Doise, W. (1978). Groups and individuals. Explanations in Social Psychology, Cambridge:
Cambridge University Press.
Multiple categorization 23
Ensari, N., & Miller, N. (1998). Effect of affective reactions by the outgroup on preferences
for crossed categorization discussion partners. Journal of Personality & Social Psychology, 75, 1503-
1527.
Ensari, N., & Miller, N. (2001). Decategorization and the reduction of bias in the crossed
categorization paradigm. European Journal of Social psychology, 31, 193-216.
Fiske, S. T., Lin, M., & Neuberg, S. L. (1999). The continuum model: ten years later. In S.
Chaiken & Y. Trope (Eds.) Dual process theories in social psychology (pp. 231-254). New York:
Guilford.
Fiske, S. T., & Neuberg, S. L. (1989). Category-based and individuating processes as a
function of information and motivation: Evidence from our laboratory. In D. Bar-Tal, C. F.
Graumann, A. W. Kruglanski, & W. Stroebe (Eds.), Stereotyping and Prejudice: changing conceptions
(pp. 83-104). Springer-Verlag: New York.
Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation, from category-
based to individuating processes; Influences of information and motivation on attention and
interpretation. In M. P. Zanna (Ed.), Advances in Experimental Social Psychology (Vol. 23, pp. 1-74).
New York: Academic Press.
Fiske, S. T., & Taylor, S. E. (1991). Social Cognition. New York: McGraw Hill.
Gaertner, S. L., & Dovidio, J. F. (2000). Reducing intergroup bias: The Common Ingroup Identity
Model. Philadelphia, PA: The Psychology Press/Taylor & Francis.
Hamilton, D. L. (1979). A cognitive attributional analysis of stereotyping. In L. Berkowitz
(Ed.), Advances in Experimental Social Psychology (Vol. 12 pp.53-84). New York: Academic Press.
Multiple categorization 24
Hastie, R. (1981). Schematic principles in human memory. In E. T. Higgins, C. P. Herman,
& M. P. Zanna (Eds.), Social cognition: The Ontario Symposium (Vol. 1, pp. 39-88). Hillsdale, NJ:
Erlbaum.
Heilman, C. M. (1995). Sex stereotypes and their effects in the workplace: What we know
and what we don’t know. Journal of Social Behaviour and Personality, 10, 3-26.
Hewstone, M., Islam, M. R., & Judd, C. M. (1993). Models of crossed categorizations and
intergroup relations. Journal of Personality & Social psychology, 64, 779-793.
Hogg, M. A., & Abrams, D. (1990). Social motivation, self-esteem and social identity. In D.
Abrams & M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp. 28-47).
New York: Harvester Wheatsheaf.
Hornsey, M. J., & Hogg, M. A. (2000). Intergroup similarity and subgroup relations: Some
implications for assimilation. Personality & Social Psychology Bulletin, 26, 948-958.
Hutter, R. R. H., & Crisp, R. J. (in press). The composition of category conjunctions.
Personality and Social Psychology Bulletin.
Jetten, J., Spears, R., & Manstead, A. S. R. (1996). Intergroup norms and intergroup
discrimination: Distinctive self-categorization and social identity effects. Journal of Personality &
Social Psychology, 71, 1222-1233.
Jetten, J., Spears, R., & Manstead, A. S. R. (2001). Similarity as a source of differentiation:
the role of group identification. European Journal of Social Psychology, 31, 621-640.
Kenworthy, J. B., Canales, C. J., Weaver, K. D., & Miller, N. (2003). Negative incidental
affect and mood congruency in crossed categorization. Journal of Experimental Social Psychology, 39,
195-219.
Multiple categorization 25
Krueger, J., & Rothbart, M. (1988). Use of categorical and individuating information in
making inferences about personality. Journal of Personality & Social Psychology, 55, 187-95.
Levine, R. A., & Campbell, D.T. (1972). Ethnocentrism: Theories of conflict, ethnic attitudes
and group behaviour. New York: Wiley.
Maass, A., & Schaller, M. (1989). Illusory correlation and social categorization- toward an
integration of motivational and cognitive factors in stereotype. Journal of Personality & Social
Psychology, 56, 709-721.
Macrae, C. N., & Bodenhausen, G. V. (2001). Social Cognition: Categorical person
perception. British Journal of Psychology, 92, 239-255.
Macrae, C. N., Stangor, C. & Milne, A. B. (1994). Activating social stereotypes: A functional
analysis. Journal of Experimental Social Psychology, 30, 370-389.
Marcus-Newhall, A., Miller, N., Holtz, R., & Brewer, M. B. (1993). Cross-cutting category
memberships with role-assignment: a means of reducing intergroup bias. British Journal of Social
Psychology, 32, 125-146.
McGarty, C., & Penny, R. E. C. (1988). Categorization, accentuation and social judgement.
British Journal of Social Psychology, 27, 147-157.
Miller, N., & Harrington, H. J. (1992). Social categorization and intergroup acceptance:
Principles for the design and development of cooperative learning teams. In R. Hertz-Lazarowitz &
N. Miller, (Eds.), Interaction in cooperative groups: the theoretical anatomy of group learning (pp. 203-
227). Cambridge University Press; New York.
Mullen, B., Brown, R. J., & Smith, C. (1992). Ingroup bias as a function of salience, relevance
and status: an integration. European Journal of Social Psychology, 22 103-122.
Multiple categorization 26
Mummendey, A. & Wenzel, M. (1999). Social discrimination and tolerance in intergroup
relations: Reactions to intergroup difference. Personality and Social Psychology Review, 3, 158-174.
Oakes, P. J., & Turner, J. C. (1990). Is limited information processing capacity the cause of
social stereotyping? In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology
(Vol. 1, pp. 111-135). Chichester: Wiley.
Rabbie, J. M., & Horowitz, M. (1969). Arousal of ingroup-outgroup bias by chance win or
loss. Journal of Personality & Social Psychology, 13, 269-277.
Roccas, S. & Brewer, M. B. (2002). Social identity complexity. Personality and Social
Psychology Review, 6, 88-106.
Sherman, J. W., Macrae, C. N., & Bodenhausen, G. V. (2000). Attention and stereotyping:
Cognitive constraints on the construction of meaningful social impressions. European Review of
Social Psychology (vol. 11, pp. 145-175). Chichester: John Wiley & Sons Ltd.
Stangor, C., & Macmillian, D. (1992). Memory for expectancy-congruent and expectancy-
incongruent information: A review of the social and social developmental literatures. Psychological
Bulletin, 111, 42-61.
Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues, 25, 79-97.
Tajfel, H. (1982). Social psychology of intergroup relations. Annual Review of Psychology, 33,
1-39.
Tajfel, H., Flament, C., Billig, M., & Bundy, R. F. (1971). Social categorization and
intergroup behaviour. European Journal of Social Psychology, 1, 149-178.
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G.
Austin & S. Worchel (Eds.), The Social Psychology of Intergroup Relations (pp. 33-47). California:
Brooks/Cole.
Multiple categorization 27
Tajfel, H., & Wilkes, A. (1963). Classification and quantitative judgement. British Journal of
Social Psychology, 54, 101-114.
Turner, J. C. (1981). The experimental social psychology of intergroup behaviour. In J. C.
Turner & H. Giles (Eds.). Intergroup Behaviour (pp. 66-101). Oxford: Basil Blackwell.
Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Re-
discovering the Social Group: A Self-categorization Theory. Oxford: Blackwell.
Urada, D. I., & Miller, N. (2000). The impact of positive mood and category importance on
crossed categorization effects. Journal of Personality and Social Psychology, 78, 417-433.
Urban, L. M., & Miller, N. (1998). A theoretical analysis of crossed categorization effects: A
meta-analysis. Journal of Personality and Social Psychology, 74, 894 908.
van Knippenberg, A. V., & Dijksterhuis, A. (2000). Social categorization and Stereotyping: a
functional perspective. European Review of Social Psychology (vol. 11, pp. 105-144). Chichester: John
Wiley & Sons Ltd.
Vanbeselaere, N. (1987). The effect of dichotomous and crossed categorizations on
intergroup discrimination. European Journal of Social Psychology, 17, 143-456.
Vanbeselaere, N. (1996). The impact of differentially valued and overlapping categorizations
upon the differentiation between positively, negatively and neutrally evaluated social groups.
European Journal of Social Psychology, 26, 75-96.
Waldzus, S. Mummendey, A. Wenzel, M. & Weber, U. (2003). Towards tolerance:
Representations of superordinate categories and perceived ingroup prototypicality. Journal of
Experimental Social Psychology, 39, 31-47.
Multiple categorization 28
Wenzel, M. (2001). A social categorization approach to distributive justice: Social identity as
the link between relevance of inputs and need for justice. British Journal of Social Psychology, 40, 315-
335.
Wenzel, M., Mummendey, A., Weber, U., & Waldzus, S. (2003). The ingroup as pars pro
toto: projection from the ingroup onto the inclusive category as a precursor to social discrimination.
Personality and Social Psychology Bulletin, 29, 461-473.
Wilder, D. A. (1986). Social Categorization: Implications for creation and reduction of
intergroup bias. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology, (Vol. 19 pp.293-
355).
Multiple categorization 29
Endnotes
1
The dependent measures and models used were extensively pre-tested. The two models presented
were of a boat and a crown. These were originally made in a pre-testing session with a group of
students. Participants were informed that for the purposes of the pre-test the models would be
referred to as ‘model 1’ and ‘model 2’. Participants were required to rate model 1 and model 2 in
terms of originality, creativity and artistic merit. For the boat; originality t (13) = .069, p = 0.95,
creativity t (13) = -.623, p = .544, and artistic merit t (13) = .203, p = .842. For the crown;
originality t (13) =.648, p =.528, creativity t (13) = 1.27, p =.226 and artistic merit t (13) = 1.72, p =
.110. Which model was presented first as ‘model 1’ was counterbalanced across participants. The
order of presentation of the models in the pre-test had no effect on participants’ ratings of
originality, creativity or artistic merit for either model. We combined the three scales into one
overall measure for each model (boat Cronbach’s = .705; crown Cronbach’s =.655). When no
reference was made to ingroup or outgroup affiliation of the person who made the model, both
models were perceived to be equally positive, t (14) = -1.20, p = .251. We thus used these models
and the three evaluative items in the subsequent experiment.
2 One way t-tests were also computed against the scale midpoint of 0 (which would indicate no
bias). There was no evidence of bias in the control condition, t (19) = -.933, p = .363, but there was
significant ingroup favoritism in the multiple alternatives condition, t (24) = -2.19, p = .038.
3
We decided to use only the most task-relevant item from the pilot evaluative measure here to
provide the simplest test of our hypotheses. An additional pre-test was carried out to check that this
‘creativity’ item could indeed be considered a measure of positive evaluation. Fifteen participants
were simply asked to rate the concept of creativity for valence on a scale anchored with negative at -
Multiple categorization 30
5, with neutral at 0 and with positive at +5. Creativity was significantly perceived to be positive
compared to the neutral midpoint: t (14) = 8.12, p < .001, (M= 3.53).
4
Although there appears to be outgroup favoritism in the multiple unrelated alternatives condition,
this is evidently more attributable to the lowering of ingroup evaluations than to an increase in
outgroup evaluations.
5
The analysis was also computed just on the creativity item and revealed exactly the same pattern, F
(1, 52) = 2.91, p = .063.
Multiple categorization 31
Table 1.
Evaluations as a function of categorization (Experiment 1).
Task
Control Multiple unrelated alternatives
Ingroup
7.27
(.961)
5.41
(2.15)
Outgroup
5.60
(1.72)
6.24
(1.64)
Intergroup Bias
1.67
(1.67)
-.823
(3.12)
Note: Higher numbers indicate more favourable evaluations for ingroup and outgroup evaluations,
but for intergroup bias higher numbers indicate more ingroup favoring bias. Standard deviations are
shown in parentheses.
Multiple categorization 32
Table 2.
Evaluations as a function of categorization (Experiment 2).
Task
Control
Multiple
Unrelated
alternatives
Multiple
Related alternatives
Ingroup
6.45
(1.11)
5.97
(1.73)
6.74
(.904)
Outgroup
6.16
(1.18)
6.62
(1.59)
6.35
(1.36)
Intergroup Bias
.249
(1.26)
-.650
(1.38)
.338
(1.18)
Note: Higher numbers indicate more favourable evaluations for ingroup and outgroup evaluations
but for intergroup bias higher numbers indicate more ingroup favoring bias. Standard deviations are
shown in parentheses.